Next, we will explore segmentation in a lossy link. Consider transferring a file of length L$=12\mathrm{.}5$ MB from Host A to Host B$.$ The path from A to B is a single link with transmission rate R $=2\mathrm{.}5$Gbps on which packets can be dropped randomly $($not due to congestion$).$ If a packet loss happens, the sender can sense the loss at the end of the transmission period for the lost packet and retransmit immediately. We will assume each packet $($regardless of size$)$ has a $2\%$ chance of being dropped on each attempt. We will compare segmentation into packets of length $2000$ B $($assume no headers need to be added$),$ vs no segmentation $($single transfer of $12\mathrm{.}5$ MB$).$

$[3$ points$]$ With segmentation, what is the expected time for the whole file to be transmitted? You don$$t need to give the exact result, but please write down a formula from which the expectation can be calculated.

$[3$ point$]$ If segmentation is not used, will the expected transmission time be the same or different as with segmentation? Explain your answer $($either give a brief argument or a calculation$).$

$[6$ points$]$ Users don't have a straightforward way to assess the "expected" delay of their network. Instead, let's assume that a user does not complain at all if the transmission can be completed in less than $200\%$ of the $$ideal$$ delay $($file size$/$transmission rate$),$ and becomes very unhappy if it is $200\%$ or more. For this question, we can refer to $200\%$ as unacceptable delay. What is the probability of unacceptable delay with and without segmentation $($using the same parameters as in the previous questions$)?$ In this question $($but not the previous one$),$ we can assume that the $1$st retransmission $($the $2$nd attempt$)$ for any packet is guaranteed to be successful.

Please give your calculations for:

i$)$ Segmented, delay at least $200\%$

ii$)$ Not segmented, delay at least $200\%$

$[3$ points$]$ Recall that one cause of loss is overutilizing the network, which is not possible in the circuit$-$switched setting. In a circuit$-$switched setting, loss is not possible. Hence, we can view $$unacceptable$$ delay as a consequence of using packet$-$switching. Based on your answer for expected delay and unacceptable delay, does segmenting get us closer to or further from ideal circuit$-$switched performance? Explain in a few sentences.

$[2$ point$]$ In the context of layered architecture, explain how segmentation relates to TCP and UDP.